Colon and gastric cancers are the third and fifth cancer types with the poorest survival. Surgery is considered the primary treatment option, which can be curative. However the decision as to whether chemotherapy administration after surgery is needed, is critical for especially stage 2 colon cancer; since a group of patients who receive chemotherapy do not have significantly improved clinical outcome. For this purpose, clinical risk factors are currently evaluated to determine patients with a high risk of progression. However this is not standardized by guidelines yet. Therefore, in this thesis my first aim was to ex vivo validate two novel independent mRNA based biomarkers, ULBP2 and SEMA5A, which have been previously identified in our lab, and to generate a prognostic signature which could stratify colon cancer patients with differential prognostic profiles using the best stratification method. I showed that a 3-group prognostic signature, SU-GIB, based on expression of these two genes is associated with cancer-specific, disease-free, and overall survival independent of clinical confounding factors in colon cancer. I performed in silico analysis in order to understand the biological details of the prognostic distinction, and revealed that patients with poorer prognosis show higher expression of pro-inflammatory cytokines and a more mesenchymal profile. Patients with better clinical outcome exhibit a more epithelial profile with higher levels of phosphorylated EGFR and Shc proteins. Analysis of high-throughput drug cytotoxicity databases also showed that colon cancer cell lines with „Bad‟ SU-GIB signature are more sensitive to a dual PI3K-MTOR inhibitor, BEZ235. In this thesis, my second goal was to define prognostic and molecular sub-groups for gastric cancer and characterize the specific biology of each sub-group. Utilizing an unsupervised approach on publicly available microarray data, I identified 3 biological groups, which harbor differential characteristics related to ECM involvement, EMT, proliferation and cell cycle. Moreover I identified distinct prognostic groups which are related to this molecular classification which can further stratify patients with known pathologic subtypes (diffuse and intestinal). I found that EMT was an important parameter for molecular and prognostic classifications for both types of cancers. I therefore, studied high-throughput drug cytotoxicity databases in order to identify selective compounds with selective growth inhibition on epithelial or mesenchymal cancer cells. I identified EGFR inhibitors as being significantly more effective on epithelial cancer cells regardless of the tissue type. My future plans include the large scale validation of SU-GIB, ex vivo validation of the gastric prognostic signature, and in vitro studies that would demonstrate effectivity of EGFR inhibitors on epithelial cancer cells and combination of EGFR inhibitors with MET inducers.